BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Interpreting Document Collections Using Topic Models - Nikos Aletr
 as\, UCL
DTSTART:20150213T120000Z
DTEND:20150213T130000Z
UID:TALK56086@talks.cam.ac.uk
CONTACT:Tamara Polajnar
DESCRIPTION:Topic models are a set of statistical methods for interpreting
  the contents of document collections. These models automatically learn se
 ts of topics from words frequently co-occurring in documents. Topics learn
 ed often represent abstract thematic subjects\, i.e Sports or Politics. To
 pics are also associated with relevant documents. These characteristics ma
 ke topic models a useful tool for organising large digital libraries. Henc
 e\, these methods have been used to develop browsing systems allowing user
 s to navigate through and identify relevant information in document collec
 tions by providing users with sets of topics that contain relevant documen
 ts. \n\nThe aim of this talk is to present methods for post-processing the
  output of topic models\, making them more comprehensible and useful to hu
 mans. First\, we look at the problem of identifying incoherent topics. We 
 show that our methods work better than previously proposed approaches. Nex
 t\, we propose novel methods for efficiently identifying semantically rela
 ted topics which can be used for topic recommendation. Finally\, we look a
 t the problem of alternative topic representations to topic keywords. We p
 ropose approaches that provide textual or image labels which assist in top
 ic interpretability. We also compare different topic representations withi
 n a document browsing system.
LOCATION:FW26\, Computer Laboratory
END:VEVENT
END:VCALENDAR